Somatosensation in Social Perception

REVIEWS

Somatosensation in social perception

Christian Keysers*‡, Jon H. Kaas§ and Valeria Gazzola*‡ Abstract | The discovery of mirror neurons in motor areas of the brain has led many to assume that our ability to understand other people’s behaviour partially relies on vicarious activations of motor cortices. This Review focuses the limelight of social neuroscience on a different set of brain regions: the somatosensory cortices. These have anatomical connections that enable them to have a role in visual and auditory social perception. Studies that measure brain activity while participants witness the sensations, actions and somatic pain of others consistently show vicarious activation in the somatosensory cortices. Neuroscientists are starting to understand how the brain adds a somatosensory dimension to our perception of other people.

Vicarious activation One of the most exciting developments of the past expressions such as “Her words were really touch- Activation of a brain region that decade is the discovery that our perception of other ing” encapsulate the intuitive link between the people is normally involved in individuals involves neurons and brain areas that were around us and our sense of touch. processing the observer’s own thought to be reserved for the control of our own In this Review, we first describe the location and con- actions and sensations, but (FIG. 1) that is now activated by seeing actions and the experience of our own emotions. First, nections of the somatosensory cortices to show similar actions or sensations in it became clear that seeing or hearing other people’s that, in contrast to the early somatosensory cortices in another person. actions recruits neurons1–8 (in monkeys) and brain the central sulcus, all other somatosensory cortices regions (in humans) in the premotor and posterior (where the later stages of somatosensory processing Proprioception parietal cortices that are also involved in programming take place) receive direct input from areas known to The sense through which we 9–18 perceive the position and similar actions . This has led to the idea that under- have visual and auditory properties. Second, we show movements of our own body. standing the inner state of other individuals relies on that in agreement with this anatomical evidence, the implicit motor simulation — that is, the activation of higher stages of somatosensory processing in the pri- motor programs that we would use to perform similar mary somatosensory cortex (SI) and secondary soma- actions19–23. Our motor cortices seemed no longer ‘pri- tosensory cortex (SII) are activated when we perceive *Social Brain Laboratory, vate’ but a part of our social brain, processing the states other people being touched, performing an action or Department of Neuroscience, of others as if they were our own. experiencing somatic pain (BOX 1). We suggest that these University Medical Center More recently, it has been suggested that brain areas vicarious activations of somatosensory cortices may have Groningen, A. Deusinglaan 2, 9713AW Groningen, involved in emotion processing, including the anterior the unique potential to provide a somatic dimension to The Netherlands. insula and the rostral cingulate cortex (rCC), might our perception of other people’s experiences. ‡Social Brain Laboratory, perform an ‘emotional simulation’ of other individuals’ Netherlands Institute for experiences, showing activity not only when we expe- Anatomy of the somatosensory system Neuroscience, Royal rience positive and negative emotions but also when Somatosensation involves the processing of tactile, Netherlands Academy of Arts 24–45 proprioceptive nociceptive and Sciences (KNAW), we witness those of others . Electrostimulation and information. Here, we Meibergdreef 47, 1105BA of similar regions of the insula causes measurable briefly describe the cortical regions involved in this Amsterdam, The Netherlands. changes in the body (for example, gastric contrac- processing and examine which of these brain regions §Department of Psychology, tions) and induces the perception of changes in the receive the visual and auditory information that would Vanderbilt University, PMB 46 (FIG. 1) 407817, 2301 Vanderbilt body . This suggests that emotional simulation is be necessary to trigger vicarious responses . Place, Nashville, Tennessee not purely conceptual but involves representations of Traditionally, in humans and monkeys the term ‘som- 37240-7817 USA. the body. Current models of social cognition there- atosensory cortices’ refers to the anterior parietal cortex Correspondence to C.K. fore incorporate the notion that motor and emotional and the upper bank (operculum) of the lateral sulcus, and V.G. e-mails: brain regions can contribute to our perception of oth- which process tactile, proprioceptive and nociceptive [email protected]; [email protected] ers by simulating other people’s actions and emotions. information. The term ‘somatosensory system’ refers to doi:10.1038/nrn2833 The somatosensory cortices have so far been ignored all of the brain regions involved in processing somato- Published online 6 May 2010 by mainstream simulation models19,20,23. However, sensory information, and includes the somatosensory

a rCC CS c PCS BA2 PF, PFG and VIP M1

ex BA BA BA BA ex rt 3a 3b 1 2 rt co co 1 PM VIP PM M1 BA BA3b SII PF BA3a PFG rCC culum S Oper ST Insula Lateral Visual sulcus cortex (opened) Thalamus Insula Thalamus STS Auditory cortex ain stem Br

b CS PCS No vicarious Strong vicarious M1 BA1BA2 BOLD responses BOLD responses

BA3b

BA3a

Figure 1 | Cortical processing networks for somatosensation. a | Posteriolateral view ofNa a humanture Re viebrainws |showing Neuroscienc arease and regions involved in somatosensation. The central sulcus (CS), posterior central sulcus (PCS) and lateral sulcus have been opened to show the areas within. The superior temporal sulcus (STS) is not opened, but it contains multisensory areas on its banks. Other sulci are not shown. For clarity, the rostral cingulate cortex (rCC) is indicated on the medial wall of the opposite hemisphere. The anterior parietal cortex includes the four strip-like areas of Brodmann area 3a (BA3a), BA3b, BA1 and BA2 (shown in green) in a rostrocaudal sequence. Regions of interest in the posterior parietal cortex include the ventral intraparietal area (VIP) and more lateral regions (PF and PFG). The lateral parietal cortex is inside the lateral sulcus. It includes the operculum of the upper bank, which contains the secondary somatosensory cortex (SII, composed of the secondary somatosensory cortex proper (S2) and the parietal ventral region (PV)), and the insula, which has regions for processing touch and pain. Somatosensory networks ultimately involve the primary motor (M1) cortex and the dorsal and ventral premotor (PM) areas for action. The thalamus (dashed grey lines) lies deep in the forebrain. b | Schematic representation of the CS, the PCS, the M1 cortex and the anterior parietal areas BA3a, BA3b, BA1 and BA2. c | Brain regions involved in somatosensation, their main connections and their source of visual input. Functional MRI experiments suggest that vicarious activation occurs in some somatosensory brain regions; the strength of the blood oxygen level-dependent (BOLD) response in these regions to observing the touch, actions or pain of others is indicated by the colour. In addition, there are brain regions with neurons that respond to the presentation of visual or auditory stimuli that also show vicarious activation; these are the PM, STS and areas PF, PFG and VIP (shown in blue).

cortices proper plus the insula and the rostral cingulate + BA2 when it is unclear to which of these subregions a cortex, which are thought to process the affective value statement applies or when it applies to all four. of somatosensory stimuli47. BA3a receives proprioceptive information, largely The anterior parietal cortex consists of four parallel, from muscle spindle receptors through the ventropos- mediolateral strips of cortex: the classical cytoarchitec- terior superior nucleus (VPS) of the thalamus, and has Nociception tonic areas 3a, 3b, 1 and 2 of Brodmann and the Vogts. close anatomical connections with the motor cortex. The sense through which we In humans, Brodmann area 3 (BA3a and BA3b) roughly BA3b is the primary area for tactile processing and perceive damage caused to our correspond to the posterior bank of the central sulcus, receives its major activating inputs from neurons in the own body — for example, by excessive heat, cold or BA1 to the crown of the postcentral gyrus and BA2 to ventroposterior nucleus (VP) of the thalamus. BA3b physical injury. the anterior bank of the postcentral gyrus (FIG. 1a,b). All also receives input from small neurons in the VP and four areas were once considered to be parts of a single in the ventroposterior inferior nucleus (VPI) that are Muscle spindle receptors ‘homunculus’, a systematic representation of the contra- activated by a broad range of stimulation intensities Receptors in the muscles that lateral body surface known as ‘SI’. It is now known that (including touch in the nociceptive range) and from measure changes in muscle 48 length and hence changes in each of the four areas constitutes a separate representation nociceptive neurons in the spinal cord and brain stem . 47 the location of the relevant with different connections and functions . Accordingly, BA1 receives strong activating inputs from BA3b and body part. the term ‘SI’ is now used to refer to BA3a + BA3b + BA1 thus is thought to be involved in a secondary cortical

Box 1 | Which studies find activity in somatosensory cortices? To identify the brain regions that are involved in observing the actions and pain of other individuals, we performed a review of the most relevant studies (see tables). First, to identify the brain regions involved in perceiving actions, we only included functional MRI studies that assessed both the perception and the execution of goal-directed hand actions in the same participants and that performed whole-brain analyses. Studies in which actions were observed for the purpose of imitation were excluded; if action imitation is a goal, observation may include motor planning, which would artificially increase the involvement of premotor regions (the dorsal premotor cortex (dPm) and ventral premotor cortex (vPm)) and, through forward models, somatosensory brain regions. The first table summarizes which brain areas were active during both action execution and action perception and reveals that Brodmann area (BA2) was consistently activated during the execution and observation of other people’s actions.

Method* BA3a BA3b BA1 BA2 SII dPM vPM PF/PFG Refs Observing an action T ? ? ? ++ + ++ ++ ++ 9 V – – – ++ + ++ ++ ++ 16 V – – – + – ++ + ++ 10 A + + + ++ + ++ ++ ++ 12 A + + + ++ + ++ ++ ++ 11 A + + + ++ + ++ ++ ++ 14 Hearing an action A – + + ++ ++ ++ ++ ++ 17 V – – ? ++ ++ ++ ++ ++ 18 ++, regions that were significantly activated by both action observation (or action sounds) and action execution and that include a local maximum; +, regions that had significant activation but were located at the fringe of larger clusters that were centred elsewhere; –, no significant activity found using methods V or A;?, activation tables were not detailed enough to determine activity of the region using method T. *Method used to determine the involvement of a brain area. A, analysis of the Statistical Parametric Mapping (SPM) Anatomy Toolbox or the Juelich Histological Atlas to determine from statistical parametric maps the regions containing voxels that were active during both action observation and execution; T, coordinates provided by published activation tables entered into the SPM Anatomy Toolbox; V, visual inspection of the published figures (BA2 is considered to be the anterior bank of the postcentral sulcus, BA1 the postcentral gyrus’s crown and BA3 the posterior bank of the central sulcus).

Only a small number of studies have measured brain SI SII AI rCC Refs activity in participants who Seeing arbitrary cues signalling that someone is in pain both experienced pain and – – + + 28* perceived another individual’s pain. To identify – + + + 32* brain regions involved in the Seeing the faces of people in pain perception of other people’s pain, we also incorporated – – + + 29* studies that included only a – – + + 45 ‘pain observation’ condition. – – + + 39* We used the activation tables in the published manuscripts – – + + 38 in combination with the SPM Seeing hands (or feet) in painful situations‡ Anatomy Toolbox to verify that the primary – – + + 44* somatosensory cortex (SI), the + ? + + 31 secondary somatosensory + + + + 30 (SII), the anterior insula (AI) and the rostral cingulate + + + + 12 cortex (rCC) are involved in + + + + 13* perceiving other people’s pain. The second table + + + + 37 summarizes the results of + + + + 40 this analysis and reveals that + + + + 41 experiments in which participants observe specific +, the paper reports a peak of activation in this region in at least one of the conditions in which body parts being harmed participants perceived the pain of others; –, a peak in this region is not reported in any of the conditions; ?, it is unclear whether this region is activated. *This study measured brain activity consistently report activity in participants who both experienced pain and perceived another individual’s pain. ‡For in SI. example, being picked by a needle.

stage of tactile processing. BA2 receives inputs from is relayed to more anterior sectors of the insula, where BA3a, BA3b and BA1 and therefore constitutes a third it is integrated with inputs from the frontal lobe, from level of cortical processing of tactile and proprioceptive all sensory modalities and from limbic structures62,66. information49. This tactile information is combined with The rCC receives nociceptive input from more lateral proprioceptive inputs from the VPS. Neurons in BA2 are nuclei in the thalamus (posterior part of the ventrome- especially responsive when objects are actively explored dial nucleus (VMpo), ventrocaudal part of the medial or manipulated with the hands so that tactile and pro- dorsal nucleus (MDvc), parafascicular nucleus (Pf) and prioceptive afferent information is combined in a proc- centrolateral nucleus (CL)) and integrates this informa- ess that we will term haptics50. Notably, BA2 also receives tion with highly processed information from various callosal connections from BA2 of the other hemisphere, cortical areas63. and these connections enable some neurons to respond This summary of the anatomy of the somatosensory to stimuli on both the ipsilateral and the contralateral system allows us to draw a number of conclusions (FIG. 2). hand during bimanual exploration51,52. The connections First, tactile and proprioceptive inputs from the thalamus between areas of anterior parietal cortex are reciprocal in are initially segregated in separate areas of the anterior that BA2 projects back to BA1, BA3b and BA3a. parietal cortex but are then combined in BA2 and sent Importantly, BA2 also has direct, reciprocal con- to SII, enabling haptics. Second, nociceptive informa- nections with regions of the fundus of the intraparietal tion from the thalamus is more broadly distributed to sulcus (such as the ventral intraparietal area (VIP)) and the cortex, with parts of the insula and rCC specifically the inferior parietal lobule (areas PF and PFG in particu- involved in processing the affective value of nociceptive lar), which combine visual, auditory and somatosensory information. Third, BA2, the SII complex, the insula and information49,53–55. In monkeys, some cells in VIP the rCC receive direct inputs from regions of cortex that respond both when the monkey is touched and when are responsive to visual and auditory stimuli, whereas it sees another individual being touched in a similar BA3b and BA1 have access to such information only way56, whereas some neurons in PF and PFG respond indirectly through BA2 and SII. Finally, the anatomical both when the monkey performs a goal-directed action connections suggest that visual or auditory information and when it sees another individual perform a similar would be most likely to influence tactile processing in action7. Moreover, these regions are thought to constitute SII, haptic processing in BA2 and SII, and nociceptive the main source of visual and auditory information for mir- processing in BA2, SII, the insula and rCC. BA3a, BA3b ror neurons in the premotor cortex57. The fact that these and BA1 should have, at best, only weaker responses regions also project to BA2 makes it plausible that BA2 to visual or auditory information through feedback could be activated vicariously when one observes the connections from SII and BA2. goal-directed actions of others. From SI, somatosensory information is sent to SII; these connections are recipro- SII: vicarious tactile processing cal, allowing areas involved in early processing stages to be Evidence for a role of somatosensory cortices in the per- influenced by areas involved in later processing stages. ception of other people’s somatosensory states originated In monkeys and humans58, SII, which lies on the pari- from a functional MRI (fMRI) study. Participants in a etal operculum (OP), can be divided into two subregions scanner were touched on their legs by an experimenter termed S2 and PV (the parietal ventral area), which cor- and then were shown movie clips of other people’s legs respond to distinct architectonic fields, OP1 and OP4, being touched by a rod or, as control stimuli, movies of respectively59. S2 and PV receive inputs from all four the same rod moving too far away from the same legs to areas of SI and are therefore involved in a third or fourth be able to touch them. Being touched activated the leg level of processing. They also have similar afferent and representations in both SI and SII. Importantly, viewing efferent cortical connections60, including connections other people being touched (as compared to the control with neighbouring cortical regions of the OP and with condition) also activated SII (but not SI; FIG. 3). SII was a number of brain regions containing cells that respond even activated when participants watched objects (for to visual and auditory input. The latter regions include example, rolls of paper) being touched compared to mov- the PF, PFG and VIP54,55 (which also provide input to ies in which the objects were not touched67. Other studies BA2), secondary auditory areas that are also responsive also showed SII activity in participants who observed the to somatosensory stimuli61, and the insula62. hands68,69 or the neck and face70 of other people being For nociception, SI and SII are thought to process the touched in movie clips. One study replicated the SII sensory discriminative aspects (that is, the intensity and activation in response to seeing objects being touched68, location) of pain63. This occurs in parallel with the affec- although another study did not70. The discovery that SII tive and motivational processing of nociceptive input, responds to the sight of humans and, sometimes, objects which is thought to take place in the insula and the ros- being touched and the fact that the neurons in SII have tral cingulate gyrus63. The posterior insula receives tha- very large receptive fields71 suggest that vicarious acti- lamic input associated with the spinothalamic pathway64 vation in SII could convey a simulation of the quality Haptics and cortical input from adjoining and nearby cortical of touch one would experience if one were touched in The sense through which we areas. Different sectors of the posterior insula seem to a similar way, rather than the precise body location at perceive the world by actively exploring it with our body — be involved in the appreciation of pain, temperature, which the touch occurred. Interestingly, watching tac- 65 for instance, finding our keys itch and pleasant touch , but they do not receive pro- tile stimulation of more erogenous zones of the body in among a pocketful of coins. nounced auditory or visual input62. This information pornographical movies also activates SII vicariously72–74.

resulted from a deliberate human action with the brain b Haptics a Passive touch response to watching movie clips in which touch was Tactile information Tactile information accidentally delivered by an object that was moved by the wind68. The study showed that although activity in SI remained subthreshold, in all cases it was stronger BA3b BA3b in the deliberate condition. This suggests that BA2 or BA1 activation relates to the toucher rather than to the BA1 BA1 sensations of the person being touched (see below). Proprioceptive Electroencephalography (EEG), which has higher BA2 information temporal resolution than fMRI, has also been used to A/V input examine which stages of somatosensory processing are sensitive to visual input75. The median nerve of par- A/V input SII SII ticipants was electrostimulated to provide a precisely timed somatosensory input, and the resulting sensory evoked potentials (SEPs) were measured on the scalp. Components of these SEPs within 40 ms of stimula- c Nociception Nociceptive information tion reflect subcortical and BA3 activity, whereas later components reflect activity in BA1, BA2 or SII76. SEPs from participants watching movie clips of a hand being touched by a cotton swab showed that components associated with BA3 were not influenced by this A/V input rCCPosterior insula BA3b visual stimulus, whereas later components (for exam- A/V input Anterior insula BA1 ple, the P45) associated with BA1, BA2 or possibly SII were influenced75. BA2 Together, the above data clearly show that, as expected from anatomical considerations (FIG. 2a), SII can be vicar- A/V input iously recruited by seeing other people being touched. It seems to be activated when we see humans or objects SII being touched, possibly conveying a feeling of what it would be like to be touched in a similar way. However, despite this vicarious SII activation, we are not usually Affective Sensory confused about who is being touched. The fact that BA3a processing processing and BA3b are only recruited when we ourselves are being touched could account for this. The role of the interme- Figure 2 | Audio/visual input to tactile, haptic and nociceptive processing. Nature Reviews | Neuroscience Although somatosensory receptors in the body provide the primary input to the diate processing stages that take place in BA1 and BA2 somatosensory system (shown in green), the processing streams for tactile, haptic and remain to be further explored but, as we will see below, 75 nociceptive information all receive input from brain regions in which cells respond to the intensity of the perceived touch and the presence audio/visual (A/V) stimuli (shown in blue). The anatomical level at which such A/V of actions in the stimuli11,17,68 could be important factors information enters these streams determines where vicarious somatosensory responses in determining their recruitment. might occur. a | The main pathway for the processing of tactile information resulting from To establish the degree to which SII contributes to passive touch first receives direct input from brain regions with A/V responses at the social perception, it will be necessary to investigate level of the secondary somatosensory cortex (SII). b | The pathway that integrates tactile how virtual (transcranial magnetic stimulation (TMS)- and proprioceptive information when we are actively manipulating an object (haptics) induced) or neurological lesions in SII change people’s receives A/V input in Brodmann area 2 (BA2) and SII. c | Nociceptive information is capacity to accurately perceive the sensations of others. distributed along multiple, parallel streams (important interconnections exist but are 77 omitted for clarity). Two of these streams are thought to primarily process the affective A certain type of synaesthesia , however, suggests a link aspects of pain, and the third stream is thought to primarily process the sensory aspects between SII activity and the way we perceive these sensa- of being hurt. Each of these streams receive A/V input at the level of the rostral cingulate tions; about 1% of people experience a vivid sensation cortex (rCC), anterior insula, BA2 and SII. of touch on their own body when they see the body of another being touched78. This effect is so automatic that these so-called ‘mirror-touch synaesthetes’ often mis- report the location on which they are touched if they The above studies showed that, in contrast to SII, simultaneously see another person being touched77,78. In SI is activated in some cases69,70 but not in others67,68 an important experiment, Blakemore et al.70 measured and BA3 is never activated during the observation of brain activity in one such synaesthete and found that she touch. BA2 and BA1 were only activated if the stimuli differed from controls in that her SI (probably including showed a human hand delivering the touch70 or when BA3) and SII regions were activated more strongly than the task focused attention on the action of touching69. In those of controls when seeing movies of other people Median nerve both cases, the hand representation in SI was activated being touched70. Interestingly, mirror-touch synaesthetes A nerve running through the 77 carpal tunnel that innervates (z coordinates >40) even when the stimuli showed a face also score higher on empathy questionnaires . Together, 70 one half of the hand being touched . Only one study has compared the brain this suggests that the degree of vicarious activation in and forearm. response to watching movie clips in which the touch somatosensory brain regions — and in particular the

a neuron system may have tight functional links with the somatosensory cortices. This begs a question that has so far not been investigated systematically: do SI or SII also contain (mirror) neurons that are active during both the observation and the execution of actions? Intriguing preliminary evidence for this possibility stems from a neuroimaging experiment showing that monkeys activate SI and SII when they are grasping an object and when they are watching another individual performing the same action86–88. –30 To test whether SI and SII might play a part in action perception in humans, we scanned participants while b they viewed objects being manipulated and while they manipulated similar objects themselves11. We found that ‘shared’ voxels (that is, voxels that were active during both observation and execution of goal-directed actions) were not restricted to the ventral premotor cortex and the posterior parietal lobe; the somatosensory cortex, BA2 in particular, contained more shared voxels in more participants than the ventral premotor cortex (FIG. 4). SII also contained shared voxels (albeit fewer than BA2), as did the dorsal premotor cortex89. A review of six studies that have examined action observation and execution using –24 fMRI (BOX 1) confirmed that BA2 is consistently active Figure 3 | Vicarious tactile activity in the secondary somatosensory cortex. Seeing during action observation (as consistently as the ventral a leg or a hand being touched elicits activity in the secondaryNa somatosensoryture Reviews | Neur cortexoscienc (SII)e premotor cortex). In contrast to BA2, more-anterior that overlaps with the activity that occurs when the participant is being touched on the sectors of SI are rarely and only weakly recruited dur- leg or hand, respectively. The left column shows a still frame from the movie clips used by ing the observation of other people’s actions. Compared Keysers et al.67 (a) and Ebisch et al.68 (b). The right column shows brain regions that are with the observation of passive touch, SII is more weakly activated exclusively by the experience of touch (shown in red) and by watching the recruited during action observation. Hearing the sound movie clips (shown in blue). The overlaps between brain areas indicate areas that are of other people’s actions also strongly activates BA2 and, activated by both the experience of touch and the observation of someone else being to a lesser extent, SII17,18. touched (shown in white). The white numbers refer to the y coordinates in the MNI (Montreal Neurological Institute) space of the coronal slice. However, findings that the same region shows increases in fMRI signals during both the observation and execution of actions (BOX 1) suggest, but do not guar- antee, that the same neurons in these voxels are involved involvement of BA3 — can determine what it feels like in both cases; distinct populations of neurons could be to see other people being touched and, more generally, involved in observation and execution but be located the vividness with which one might empathically share within the same fMRI voxels11, or changes in fMRI signal other people’s physical experiences. could originate from the same neurons but only reflect subthreshold synaptic input during perception. The same BA2: vicarious haptics and proprioception caveat applies to the cases of touch and nociception (see SI is not just essential for our sense of touch — in below). Systematic single-cell recordings are needed to humans, lesions to SI lead to devastating impairments investigate whether the same neurons in the somatosen- in motor control79, and in monkeys, deactivation of BA2 sory cortices are involved in both action observation and impairs the motor control needed for grasping80. Does SI execution, and what aspect of an action is represented also help us to perceive the actions of others? The study in these putative visually and auditorially responsive of the cortical processing of other individuals’ actions somatosensory neurons. Nevertheless, two sources have has been dominated by the discovery of mirror neurons provided evidence that some of the same somatosensory in monkeys. These neurons, which respond both when neurons are active during action observation and execu- a monkey performs an action and when it sees or hears tion. First, the only experiment that has examined the another individual perform a similar action1–8, have so presence of mirror neurons in humans using repetition Somatosensory evoked far been reported in regions involved in motor planning: suppression and a whole-brain analysis found repeti- potentials 1,3–6,8 Electroencephalographic (EEG) the ventral premotor cortex and the posterior pari- tion suppression in SI during both the observation and 2,7 16 signals recorded from the scalp etal cortex (areas PF and PFG and the anterior intra- the execution of actions . Second, the shallow depth of that are induced by the parietal sulcus3). Consequently, most theoretical papers recording of some of the neurons that seemed to exhibit repeated application of a regarding the function of mirror neurons focus on the mirror properties in the anterior bank of the intrapari- somatosensory stimulus to the motor (as opposed to the somatosensory) side of action etal sulcus3 suggests that they may have been located in body or by electrically 19,20,23,81–84 triggering activity in the simulation . However, the finding that half of BA2 (A. Iriki, personal communication). somatosensory fibres in the neurons in the ventral premotor cortex also respond As mentioned above, a limitation of fMRI studies peripheral nerves. to somatosensory stimulation85 suggests that the mirror is that activity in a particular brain region during both

a Parietal (motor with fMRI) to the representation of a particular body or somatosensory?) part. Hence, it is possible to be confident that the obser- Premotor (motor) vation of hand actions specifically triggers the repre- dPM PF/PFG sentation of hand actions in BA2 (BOX 1). Additionally, executing hand and mouth actions causes activity in Temporal vPM (visual) dorsal and ventral SI regions, respectively, and perceiving mouth and hand actions triggers vicarious activity in the corresponding locations17. Multi-voxel pattern classification of activity data in SI during action perception can identify the body part that was used for an action performed by another individual90. Finally, watching movies of right-hand grasping activates the left BA2 more than the right BA2, and watching mov- MTG ies of left-hand grasping activates the right BA2 more than the left BA2 (REF. 91). Together, these data suggest that vicarious BA2 activity could provide fine-grained, somatotopically specific representations of other Masking with BA2 people’s actions.

b What does vicarious activity in BA2 convey? A number BA2 of studies help us to understand what perceptual content vicarious BA2 activity could convey. First, observing hand movements that involve extreme joint stretching (for example, stretching beyond the normal physiological angle) activates BA2 strongly92, and deactivating BA2 using TMS reduces motor-evoked potentials in the hand when seeing such extreme joint stretching93. Second, BA2 is more, or sometimes only, active when viewing hands manipulating objects (for example, grasping a cup) compared with actions that do not involve objects (for example, pointing or mimicked grasping)94,95. Additionally, viewing someone move a heavy object activates BA2 more strongly than viewing someone moving a light object96. Figure 4 | Vicarious activity during the observation of Together, and in accordance with anatomical con- actions. a | Shared voxels, thatNa is,tur voxelse Revie thatws show| Neur activityoscience siderations (FIG. 2b), these data suggest that BA2 might during both action observation and action execution, have be particularly involved in vicariously representing the been reliably identified in four large clusters of brain areas: haptic combination of tactile and proprioceptive sig- the dorsal premotor cortex (dPM) and ventral premotor nals that would arise if the participant manipulated the cortex (vPM), which are involved in motor control; the object in the observed way (FIG. 2b). This conclusion is posterior mid-temporal gyrus (MTG), which is involved in confirmed by the observation that when participants visual perception; and a large cluster encompassing watched a film in which the actors were seen to manip- multiple regions of the parietal lobe. This cluster has traditionally been considered to be part of the posterior ulate objects, vicarious activity in SI was consistently 97 parietal lobe because it encompasses areas PF and PFG, in observed . Interestingly, the more motor expertise which mirror neurons have been recorded in monkeys, thus people have, the more they activate BA2 when observ- associating this area with motor control. b | However, ing actions related to their field of expertise — even if showing only those voxels of a that fall within Brodmann they are not directed at objects (for example, dancers area 2 (BA2) reveals how much of the parietal shared voxels watching a dance performance98–101). This suggests that are actually part of the somatosensory cortex. This although haptic object manipulation might be the opti- indicates that activity in this part of the cluster probably mal stimulus for BA2, BA2 might underlie our capacity represents vicarious haptic activity instead of vicarious to know what it would feel like to move one’s body in an motor activity. The figure is based on data from REF. 11. observed way more generally. The consistency with which BA2 is activated during the observation of actions additionally sheds light on the the execution and perception of an action does not nec- presence of BA2 activity in only some of the touch- essarily indicate that both scenarios induce activity in observation studies mentioned above68,69. Viewing touch neurons representing the same information (for exam- activates BA2 more when touch is delivered by a human ple, a particular haptic sensation or motor program) — hand than by an object68. Focusing attention on the act that is, the information being processed might not be of touching (for example, following instructions to count the same. However, the somatotopic organization of SI the number of touching actions)69 activates the hand allows one to link the location of activity (measurable area of BA2 more than focusing attention on the touched

individual (that is, when the touched individual or of others, premotor, posterior parietal and somatosen- object but not the toucher is shown)67. Together, these sory vicarious activity each contribute primarily to considerations propose a functional complementarity a different aspect of the perception of other people’s in vicarious somatosensory activity, with BA2 relating actions: programming our motor response, sensing the to the ‘sharing’ of the haptic aspects of actions and SII intentions of others and experiencing what it would activity to the sharing of passive touch. feel like to move one’s own body in the observed way, In contrast to most reviews on the neural mecha- respectively. In support of this idea, SI activity is cor- nisms underlying action observation19,20,23,81–83,102, we related with the accuracy with which a person judges conclude that the observation of other people’s actions how another person feels111. recruits not only the ventral premotor cortex and posterior parietal cortex involved in programming the SI and SII: vicarious nociception observed actions, but also BA2 and to a lesser extent We have all experienced that witnessing the pain of oth- SII, which are involved in sensing how our own body ers is aversive. If we see our partner’s face expressing would move and interact with the object in the observed intense pain, we feel deeply distressed. If we see him way. The simulation of actions would thus involve both or her cut their finger with a sharp kitchen knife, we simulating the motor output that would be necessary not only feel distress, we often feel compelled to grasp for performing the observed action and simulating the our own finger. About a third of people feel pain on haptic somatosensory input that would accompany the corresponding part of their own body when they the performance of those actions. Such a link between the see certain injuries of other people112. Neuroimaging motor and somatosensory system during action obser- research is now starting to shed light on the multifac- vation would be consistent with the link between these eted nature of this empathic pain. In brief, this research two systems during action execution, in which the shows that if all we know is that another person is in expectation of touch is a fundamental component of pain, we vicariously recruit brain regions involved forward models in goal-directed motor control103,104. in the affective experience of pain: the anterior insula Interestingly, one paper also shows that of two simi- and rCC28,29,32,38,39,45. Here, we will show that whenever lar actions, those that activate BA2 more strongly are our attention is directed to the somatic cause of the recognized more rapidly105. pain of others, somatosensory cortices also become Facial expressions are a special type of action. vicariously activated. This process could trigger a more Experiments that have examined the neural structures localized, somatic sensation of pain in a particular involved in both the observation and execution of body part that drives us to grab our finger in the above dynamic facial expressions are in agreement that, akin to example. This conclusion stems from the comparison observing hand actions, observing the facial expressions of fMRI experiments that used different types of stimuli of others vicariously activates ventral sectors of BA2 and/ to inform the participant about the pain of others. We or SII that are involved in sensing self-produced facial describe these studies below. expressions36,106,107. Real and virtual (TMS) lesions in In a seminal pair of fMRI studies, participants in the these somatosensory face representations impair the scanner were shown a coloured cue on a screen indi- recognition of facial expressions108,109, which suggests cating when another individual, present in the scanner that vicarious somatosensory representations of what it room, received a painful electrical shock28,32. In a second feels like to move the face in the observed way contribute set of fMRI experiments, participants were informed to the recognition of other people’s facial expressions. about the pain of others by viewing facial expressions In contrast to these findings, studies examining activity on a screen29,38,39,45. In the last group, participants viewed during the observation of pained facial expressions have images of specific body parts being hurt — for exam- not reported activity in somatosensory cortices29,38,39,45. ple, a hand being pinpricked or deeply penetrated by Under what conditions somatosensory representa- a hypodermic needle37,40–44, or a foot being hit by a 30,31 Forward model tions of the face are vicariously activated when one door . Comparing the loci of brain activation in these A system that predicts the observes other people’s facial expressions remains to be and various other fMRI experiments (BOX 1) confirms consequences of a motor explored further. previous findings that perceiving the pain of others command in sensory When we perceive the actions of others, we often vicariously activates the higher levels of affective noci- (somatosensory in become aware of their intentions and feelings and ceptive processing — that is, the anterior insula and/or particular) terms. modify our own actions accordingly. How do the rCC. The somatosensory cortices (SI and SII), however, Qualia brain areas in this putative extended somato-motor were only vicariously activated in some conditions, and (Plural of quale.) A quality or action simulation circuit contribute to these processes? almost never while people viewed painful facial expres- property as it is perceived or Electrical brain stimulation in awake surgical patients sions or abstract cues. This led the authors of one paper experienced by a person. For instance, although a tomato might provide an answer; stimulating the premotor to state: “Empathy for pain involves the affective but 28 has the same physical cortex leads to overt movements but patients firmly not sensory components of pain.” However, in the properties regardless of deny that they actually moved110. Further, stimulating eight studies in which the noxious (somatic) event whether it is seen by a typical the inferior parietal lobule creates a strong intention itself was shown to the subjects30,31,37,40–44, SI and SII were or a colour-blind viewer, the and desire to move without actual movements taking vicariously activated in all but one experiment (BOX 1). qualia it will trigger in the two 110 individuals differ substantially, place . Finally, stimulation of the somatosensory cor- Empathy for pain therefore involves the somatosensory with a ‘redness’ perception tex triggered somatosensory qualia in the hands and cortices, but only if one attends to the localized somatic triggered only in the former. body110. This suggests that when we observe the actions cause of the pain.

Some of these studies have identified further fac- the most likely candidates because they are associated tors that determine whether somatosensory cortices are with these aspects of pain perception during first-hand activated while subjects view the tactile pain of others: pain experiences63. activations in SI are stronger when participants imag- Although it is difficult to determine from the lim- ine that they are in this painful situation themselves ited details available from published activation tables than when they imagine someone else being in that which parts of SI are recruited by the observation of situation30, when they explicitly judge how painful being other people’s body parts being harmed, of the eight in that situation would be41 and when the visual stimu- fMRI studies that examined the observation of hands or lus suggests more intense pain, such as when a needle is feet in painful situations, six explicitly report vicarious shown penetrating a hand deeply37 as opposed to pricking activation in coordinates that correspond to BA2 or BA1 it44. In light of this ‘intensity coding’, it is possible that (REFS 30,31,37,40,42,43), whereas only one explicitly studies on neutral touch failed to find consistent SI activ- mentions activation in BA3 (REF. 37). Given that the EEG ity because the touch stimulus they used was less intense data also point to vicarious activations of BA1, BA2 or than the painful tactile stimuli used in pain studies that SII but not BA3 (REF. 76), it seems that, as for touch and consistently find SI activity. action, vicarious activity for somatic pain is restricted to Although these studies show that, in general, SI and the higher levels of somatosensory processing (that is, SII are activated when individuals witness the somatic regions that receive direct auditory or visual input; FIG. 2c), pain of others, substantial inter-individual differences whereas BA3 remains ‘private’, only being activated by seem to exist. When observing photographs of injuries the first-hand experience of pain. This difference could (for example, an athlete breaking his leg), about one-third again account for why observing the pain of others can of participants report feeling pain on the corresponding be touching in a localized way without causing confu- part of their own body. The remaining two-thirds report sion about who was being hurt (that is, the observer or negative feelings without a sense of somatic pain. fMRI the observed person). Further evidence for the soma- showed that vicarious SI and SII activity was triggered totopical sharing of other people’s pain stems from the by such images only in the participants who experienced study of motor-evoked potentials (BOX 2). localized vicarious pain112. This provides further evidence Taken together, these data indicate that we can that vicarious SI and SII activity adds a somatic dimension share the pain of others in two ways. If all we know is to social perception and urges us to start exploring the that the observed person is in pain, we share the affec- neural basis for such inter-individual phenomenological tive aspects of their distress through vicarious activity differences in vicarious experiences. in the anterior insula and rCC. If, however, we focus on Together, these studies suggest reasons for the varia- the somatic causes of that pain, we additionally share its tion of ‘shared pain’ from a generic distress to a specific somatic consequences by vicariously recruiting BA1, bodily feeling; only if participants directly witness an BA2 and/or SII (FIG. 2c). intense, localized, harmful somatic event as the cause for other people’s pain do they vicariously activate their Conclusions and future directions somatosensory cortices (in addition to the anterior In summary, we have seen that although the truly pri- insula and rCC). Owing to the somatotopic organization mary somatosensory cortex, BA3, is only systematically of SI, its vicarious activation is likely to add a localized, involved in processing signals that originate in our own somatosensory feeling to our empathy for pain in these body, the second (BA1) and third (BA2 and SII) corti- conditions. This idea finds further support from an EEG cal somatosensory processing stages can be vicariously study75 in which the P45 component of the SEP, which is recruited by the sight of other people being touched, thought to originate in BA1, BA2 and, possibly, SII76, is performing actions or experiencing somatic pain (FIG. 2). increased when an individual sees someone else’s hand In particular, when we see one person touching another, being deeply penetrated by a needle. Interestingly, in this BA2 seems to primarily represent the action of the study the BA1, BA2 and SII activation levels, as measured toucher, whereas SII seems to represent the tactile sen- by the P45 modulation, correlated with the intensity of sations of the individual being touched. The notion that the pain that participants attributed to the actors in the our premotor cortex is not the private fort of our own movie clips75. In accordance with the intensity-coding actions but a shared arena in which our actions and role of the somatosensory cortex during first-hand those of others can coexist should now be extended to nociception63, this suggests that vicarious somatosen- the somatosensory cortices. Much remains to be inves- sory activity also participates in conveying a quantita- tigated to understand the mechanisms and significance tive sense of pain during social perception. Importantly, of these activations. people do not always perceive the most intense pain as There are a number of promising approaches that the most unpleasant or aversive75, and the P45 modula- could be used in the study of these activations. Firstly, tion did not correlate with how unpleasant or aversive single-cell recordings could be used to examine whether they rated the pain to have been in the visual stimuli75. the same neurons in somatosensory cortices respond This suggests that our perception of how unpleasant or during the experience and observation of somatosensory aversive an experience might have been for someone else states. Secondly, multi-voxel pattern classification could depends primarily on structures other than the somato- be used to analyse data from neuroimaging studies113 sensory cortices. The insula and cingulate cortex, which in which the tactile, proprioceptive and nociceptive are active in all empathic pain experiments28–32,37–45, are content of somatosensory and visual stimuli are varied

Box 2 | Vicarious motor-evoked potentials Stimulating the hand representation in the primary motor a Pain b Static c Touch cortex using transcranial magnetic stimulation (TMS) leads to activity in muscles of the hand, the so-called motor-evoked potential (MEP). Pinpricking these muscles while TMS is applied reduces the MEP in the pricked muscle but not in neighbouring muscles of the arm118, demonstrating a somatotopically specific effect of nociceptive tactile input. Why people show reduced activity in a pinpricked muscle remains poorly understood but, interestingly, seeing a video clip of another individual’s hand being deeply penetrated by a needle while a TMS pulse is applied to the observer also reduces the MEP of the observer’s hand muscle119–122 (see the figure, part a), compared with the MEP during TMS in an observer who watches a video of a static hand (part b). This reduction is greatest in the muscle that is seen to be pricked119–122. The amplitude of this MEP reduction correlates with the participant’s ratings of the intensity of 0.5 mV the pain in the other person caused by the needle 20 ms puncture119–122. Seeing the same hand being touched does not reduce the MEP significantly (part c). Participants were additionally asked to mention whetherNatur ethey Revie thoughtws | Neur theoscienc pain e of the observed individual was restricted to the punctured muscle or spread to a neighbouring muscle while MEPs were measured from both of these muscles in the observers. Participants who reported experiencing localized pain only showed MEP reductions in the muscle that was punctured in the video clip, whereas those who thought the pain would have spread to neighbouring muscles of the observed individual showed reduced MEP in neighbouring muscles120. The fine-grained somatotopy of these effects suggests that during the observation of pain, vicarious activation may spill over from the SI to the M1 region owing to the strong and reciprocal connections between these two brain regions (FIG. 1).

systematically100 to examine the cortical areas in which these states as qualia on their own body because BA3 sub-aspects of somatosensation are represented. Thirdly, is normally excluded from vicarious processing. In fMRI studies that analyse effective connectivity could be support of this idea, synaesthetes who experience used to unravel which of the possible anatomical con- somatosensory qualia on their own body while wit- nections reviewed here actually trigger vicarious activity nessing others being touched show vicarious activity in somatosensory cortices. Finally, to disentangle what in the central sulcus (BA3)70. Reducing vicarious BA3 qualia each somatosensory cortex may convey, people’s activity in synaesthetes using TMS or other techniques reported feelings could be examined during electrostim- could be used to investigate the role of BA3 in distin- ulation of each of these somatosensory cortices, and this guishing our own states from those of others. could be combined with analysis of changes in people’s Other fields of neuroscience are also starting to rec- perception of others during manipulation of activity in ognize that visual input modulates the somatosensory these regions using TMS. cortices. For example, an elegant series of psychophysical Many open questions remain, but multiple sources of studies has shown that simply seeing a part of one’s own evidence already suggest that vicarious somatosensory body increases the accuracy with which one can localize activity has a crucial role in empathy and social percep- an invisible tactile stimulus on that body part, and that tion. More-empathic people show stronger activation this effect is likely to be due to top-down visual modula- in BA2 than less-empathic people when they perceive tion of SI neurons115. In addition, the ‘rubber-hand illu- the actions of others17, synaesthetes who feel the sen- sion’ has shown that seeing an object being brushed in sations of others activate their somatosensory cortices synchrony with somatosensory stimulation of our body more strongly than non-synaesthetes70, and lesions leads to the illusion that this object has become part of in the somatosensory cortices impair our capacity to our body116,117. These phenomena show how intimately feel the emotions of others108,109. visual and somatosensory information integrate in Our Review raises a further question: how can the the brain. somatosensory cortices process the somatic states of We hope that this Review will establish more others without us experiencing these states continuously, firmly the idiomatic idea that other people’s sensa- as qualia on our own body? The fact that BA3 (unlike tions, pain and actions can be ‘touching’ and that it higher levels of somatosensory processing) is only active will trigger new experiments aimed at further explor- during our own somatosensory experiences may be the ing this idea. As a side-product of this research, we crucial factor. Blindsight patients114, who have V1 dam- will also gain a greater understanding of why people age, process visual information in higher visual areas are willing to pay billions of dollars to watch movies, without experiencing visual qualia. Analogously, peo- from westerns95 to more frivolous genres72–74. It seems ple could process the somatosensory states of others in that movies can get under our skin as if we were the higher-level somatosensory regions without experiencing protagonists themselves.

1. Caggiano, V., Fogassi, L., Rizzolatti, G., Thier, P. & 27. Calder, A. J., Keane, J., Manes, F., Antoun, N. & 52. Killackey, H. P. , G o u l d , H . J., Cusick, C. G., Pons, T. P. & Casile, A. Mirror neurons differentially encode the Young, A. W. Impaired recognition and experience of Kaas, J. H. The relation of corpus callosum peripersonal and extrapersonal space of monkeys. disgust following brain injury. Nature Neurosci. 3, connections to architectonic fields and body surface Science 324, 403–406 (2009). 1077–1078 (2000). maps in sensorimotor cortex of new and old world 2. Fogassi, L. et al. Parietal lobe: from action 28. Singer, T. et al. Empathy for pain involves the affective monkeys. J. Comp. Neurol. 219, 384–419 (1983). organization to intention understanding. Science 308, but not sensory components of pain. Science 303, 53. Maunsell, J. H. & van Essen, D. C. The connections of 662–667 (2005). 1157–1162 (2004). the middle temporal visual area (MT) and their 3. Fujii, N., Hihara, S. & Iriki, A. Social cognition in 29. Botvinick, M. et al. Viewing facial expressions of pain relationship to a cortical hierarchy in the macaque premotor and parietal cortex. Soc. Neurosci. 3, engages cortical areas involved in the direct monkey. J. Neurosci. 3, 2563–2586 (1983). 250–260 (2008). experience of pain. Neuroimage 25, 312–319 (2005). 54. Lewis, J. W. & Van Essen, D. C. Corticocortical 4. Gallese, V., Fadiga, L., Fogassi, L. & Rizzolatti, G. 30. Jackson, P. L., Brunet, E., Meltzoff, A. N. & Decety, J. connections of visual, sensorimotor, and multimodal Action recognition in the premotor cortex. Brain 119, Empathy examined through the neural mechanisms processing areas in the parietal lobe of the macaque 593–609 (1996). involved in imagining how I feel versus how you feel monkey. J. Comp. Neurol. 428, 112–137 (2000). 5. Keysers, C. et al. Audiovisual mirror neurons and pain. Neuropsychologia 44, 752–761 (2006). 55. Rozzi, S. et al. Cortical connections of the inferior action recognition. Exp. Brain Res. 153, 628–636 31. Jackson, P. L., Meltzoff, A. N. & Decety, J. How do we parietal cortical convexity of the macaque monkey. (2003). perceive the pain of others? A window into the neural Cereb. Cortex 16, 1389–1417 (2006). 6. Kohler, E. et al. Hearing sounds, understanding processes involved in empathy. Neuroimage 24, 56. Ishida, H., Nakajima, K., Inase, M. & Murata, A. actions: action representation in mirror neurons. 771–779 (2005). Shared mapping of own and others’ bodies in Science 297, 846–848 (2002). 32. Singer, T. et al. Empathic neural responses are visuotactile bimodal area of monkey parietal cortex. 7. Rozzi, S., Ferrari, P. F., Bonini, L., Rizzolatti, G. & modulated by the perceived fairness of others. Nature J. Cogn. Neurosci. 22, 83–96 (2010). Fogassi, L. Functional organization of inferior parietal 439, 466–469 (2006). This single-cell recording study in macaques lobule convexity in the macaque monkey: 33. Carr, L., Iacoboni, M., Dubeau, M. C., Mazziotta, J. C. provides the first systematic evidence of the electrophysiological characterization of motor, sensory & Lenzi, G. L. Neural mechanisms of empathy in existence in monkeys of single neurons that respond and mirror responses and their correlation with humans: a relay from neural systems for imitation to both when the monkey is being touched and when it cytoarchitectonic areas. Eur. J. Neurosci. 28, limbic areas. Proc. Natl Acad. Sci. USA 100, sees someone else being touched. 1569–1588 (2008). 5497–5502 (2003). 57. Keysers, C. & Perrett, D. I. Demystifying social 8. Umilta, M. A. et al. I know what you are doing: a 34. Dapretto, M. et al. Understanding emotions in others: cognition: a Hebbian perspective. Tre n d s C o g n . S c i . neurophysiological study. Neuron 31, 155–165 mirror neuron dysfunction in children with autism 8, 501–507 (2004). (2001). spectrum disorders. Nature Neurosci. 9, 28–30 58. Disbrow, E., Litinas, E., Recanzone, G. H., Slutsky, D. A. 9. Grezes, J., Armony, J. L., Rowe, J. & Passingham, R. E. (2006). & Krubitzer, L. A. Thalamocortical connections of the Activations related to ‘mirror’ and ‘canonical’ neurons 35. Jabbi, M., Swart, M. & Keysers, C. Empathy for parietal ventral area (PV) and the second in the human brain: an fMRI study. Neuroimage 18, positive and negative emotions in the gustatory somatosensory area (S2) in macaque monkeys. 928–937 (2003). cortex. Neuroimage 34, 1744–1753 (2007). Thalamus Relat. Syst. 1, 289–302 (2002). 10. Filimon, F., Nelson, J. D., Hagler, D. J. & Sereno, M. I. 36. van der Gaag, C., Minderaa, R. & Keysers, C. Facial 59. Eickhoff, S. B., Grefkes, C., Zilles, K. & Fink, G. R. The Human cortical representations for reaching: mirror expressions: what the mirror neuron system can and somatotopic organization of cytoarchitectonic areas on neurons for execution, observation, and imagery. cannot tell us. Soc. Neurosci. 2, 179–222 (2007). the human parietal operculum. Cereb. Cortex 17, Neuroimage 37, 1315–1328 (2007). 37. Costantini, M., Galati, G., Romani, G. L. & 1800–1811 (2007). 11. Gazzola, V. & Keysers, C. The observation and Aglioti, S. M. Empathic neural reactivity to noxious 60. Disbrow, E., Litinas, E., Recanzone, G. H., Padberg, J. execution of actions share motor and somatosensory stimuli delivered to body parts and non-corporeal & Krubitzer, L. Cortical connections of the second voxels in all tested subjects: single-subject analyses of objects. Eur. J. Neurosci. 28, 1222–1230 (2008). somatosensory area and the parietal ventral area in unsmoothed fMRI data. Cereb. Cortex 19, 38. Decety, J., Echols, S. & Correll, J. The blame game: the macaque monkeys. J. Comp. Neurol. 462, 382–399 1239–1255 (2009). effect of responsibility and social stigma on empathy (2003). Using unsmoothed data and single-subject analysis, for pain 22, 985-997 (2009). 61. Hackett, T. A. in Evolution of Nervous Systems (ed. this paper shows that BA2 and SII are involved in 39. Lamm, C., Batson, C. D. & Decety, J. The neural Kaas, J. H.) 109–119 (Elsevier, Oxford, 2007). action observation and action execution. substrate of human empathy: effects of perspective- 62. Mufson, E. J. & Mesulam, M. M. Insula of the old 12. Gazzola, V., Rizzolatti, G., Wicker, B. & Keysers, C. taking and cognitive appraisal. J. Cogn. Neurosci. 19, world monkey. II: Afferent cortical input and comments The anthropomorphic brain: the mirror neuron system 42–58 (2007). on the claustrum. J. Comp. Neurol. 212, 23–37 responds to human and robotic actions. Neuroimage 40. Lamm, C. & Decety, J. Is the extrastriate body area (1982). 35, 1674–1684 (2007). (EBA) sensitive to the perception of pain in others? 63. Brooks, J. & Tracey, I. From nociception to pain 13. Gazzola, V. et al. Aplasics born without hands mirror Cereb. Cortex 18, 2369–2373 (2008). perception: imaging the spinal and supraspinal the goal of hand actions with their feet. Curr. Biol. 17, 41. Lamm, C., Meltzoff, A. N. & Decety, J. How do we pathways. J. Anat. 207, 19–33 (2005). 1235–1240 (2007). empathize with someone who Is not like us? A 64. Craig, A. D. & Zhang, E. T. Retrograde analyses of 14. Turella, L., Erb, M., Grodd, W. & Castiello, U. Visual functional magnetic resonance imaging study. spinothalamic projections in the macaque monkey: features of an observed agent do not modulate human J.Cogn. Neurosci. 22, 362-376 (2009). input to posterolateral thalamus. J. Comp. Neurol. brain activity during action observation. Neuroimage 42. Lamm, C., Nusbaum, H. C., Meltzoff, A. N. & 499, 953–964 (2006). 46, 844–853 (2009). Decety, J. What are you feeling? Using functional 65. Björnsdotter, M., Löken, L., Olausson, H., Vallbo, Å. & 15. Buccino, G. et al. Neural circuits underlying imitation magnetic resonance imaging to assess the modulation Wessberg, J. Somatotopic organization of gentle touch learning of hand actions: an event-related fMRI study. of sensory and affective responses during empathy for processing in the posterior insular cortex. J. Neurosci. Neuron 42, 323–334 (2004). pain. PLoS ONE 2, e1292 (2007). 29, 9314–9320 (2009). 16. Dinstein, I., Hasson, U., Rubin, N. & Heeger, D. J. 43. Morrison, I. & Downing, P. E. Organization of felt and 66. Augustine, J. R. Circuitry and functional aspects of the Brain areas selective for both observed and executed seen pain responses in anterior cingulate cortex. insular lobe in primates including humans. Brain Res. movements. J. Neurophysiol. 98, 1415–1427 (2007). Neuroimage 37, 642–651 (2007). Rev. 22, 229–244 (1996). 17. Gazzola, V., Aziz-Zadeh, L. & Keysers, C. Empathy and 44. Morrison, I., Lloyd, D., di Pellegrino, G. & Roberts, N. 67. Keysers, C. et al. A touching sight: SII/PV activation the somatotopic auditory mirror system in humans. Vicarious responses to pain in anterior cingulate during the observation and experience of touch. Curr. Biol. 16, 1824–1829 (2006). cortex: is empathy a multisensory issue? Cogn. Affect. Neuron 42, 335–346 (2004). 18. Ricciardi, E. et al. Do we really need vision? How blind Behav. Neurosci. 4, 270–278 (2004). This was the first fMRI study to show that SII is people ‘see’ the actions of others. J. Neurosci. 29, 45. Saarela, M. V. et al. The compassionate brain: humans active both when people experience touch and 9719–9724 (2009). detect intensity of pain from another’s face. Cereb. when they see other people and objects 19. Rizzolatti, G. & Craighero, L. The mirror-neuron Cortex 17, 230–237 (2007). being touched. system. Annu. Rev. Neurosci. 27, 169–192 (2004). 46. Penfield, W. & Faulk, M. E. Jr. The insula; further 68. Ebisch, S. J. et al. The sense of touch: embodied 20. Rizzolatti, G. & Fabbri-Destro, M. The mirror system observations on its function. Brain 78, 445–470 simulation in a visuotactile mirroring mechanism for and its role in social cognition. Curr. Opin. Neurobiol. (1955). observed animate or inanimate touch. J. Cogn. 18, 179–184 (2008). 47. Kaas, J. H. in The Human Nervous System 2nd edn Neurosci. 20, 1611–1623 (2008). 21. Gallese, V., Keysers, C. & Rizzolatti, G. A unifying view (eds Paxinos, G. & Mai, J. K.) 1059–1092 (Elsevier, 69. Schaefer, M., Xu, B., Flor, H. & Cohen, L. G. Effects of of the basis of social cognition. Tre n d s C o g n . S c i . 8, London, 2004). different viewing perspectives on somatosensory 396–403 (2004). 48. Craig, A. D. Retrograde analyses of spinothalamic activations during observation of touch. Hum. Brain 22. Keysers, C. & Gazzola, V. Towards a unifying neural projections in the macaque monkey: input to ventral Mapp. 30, 2722–2730 (2009). theory of social cognition. Prog. Brain Res. 156, posterior nuclei. J. Comp. Neurol. 499, 965–978 70. Blakemore, S. J., Bristow, D., Bird, G., Frith, C. & 379–401 (2006). (2006). Ward, J. Somatosensory activations during the 23. Iacoboni, M. & Dapretto, M. The mirror neuron 49. Pons, T. P. & Ka a s, J . H. Corticocortical connections of observation of touch and a case of vision-touch system and the consequences of its dysfunction. area 2 of somatosensory cortex in macaque monkeys: synaesthesia. Brain 128, 1571–1583 Nature Rev. Neurosci. 7, 942–951 (2006). a correlative anatomical and electrophysiological (2005). 24. Bastiaansen, J. A., Thioux, M. & Keysers, C. Evidence study. J. Comp. Neurol. 248, 313–335 (1986). This fMRI case study linked mirror-touch for mirror systems in emotions. Phil. Trans. R. Soc. 50. Lederman, S. J. & Klatzky, R. L. Haptic perception: a synaesthesia, — that is, the vivid experience of Lond. B 364, 2391–2404 (2009). tutorial. Atten. Percept. Psychophys. 71, 1439–1459 observed touch on one’s own body — with 25. Wicker, B. et al. Both of us disgusted in my insula: the (2009). hyperactivity during touch observation during common neural basis of seeing and feeling disgust. 51. Iwamura, Y., Tanaka, M., Iriki, A., Taoka, M. & Toda, T. touch observation. Neuron 40, 655–664 (2003). Processing of tactile and kinesthetic signals from 71. Krubitzer, L., Clarey, J., Tweedale, R., Elston, G. & 26. Adolphs, R., Tranel, D. & Damasio, A. R. Dissociable bilateral sides of the body in the postcentral gyrus of Calford, M. A redefinition of somatosensory areas in neural systems for recognizing emotions. Brain Cogn. awake monkeys. Behav. Brain Res. 135, 185–190 the lateral sulcus of macaque monkeys. J. Neurosci. 52, 61–69 (2003). (2002). 15, 3821–3839 (1995).

72. Arnow, B. A. et al. Women with hypoactive sexual 91. Shmuelof, L. & Zohary, E. Dissociation between A TMS study showing that the somatosensory desire disorder compared to normal females: a ventral and dorsal fMRI activation during object cortex contributes to recognizing the facial functional magnetic resonance imaging study. and action recognition. Neuron 47, 457–470 expressions of other people. Neuroscience 158, 484–502 (2009). (2005). 110. Desmurget, M. et al. Movement intention after 73. Ferretti, A. et al. Dynamics of male sexual arousal: 92. Costantini, M. et al. Neural systems underlying parietal cortex stimulation in humans. Science 324, distinct components of brain activation revealed by observation of humanly impossible movements: an 811–813 (2009). fMRI. Neuroimage 26, 1086–1096 (2005). fMRI study. Cereb. Cortex 15, 1761–1767 (2005). 111. Zaki, J., Weber, J., Bolger, N. & Ochsner, K. The neural 74. Hamann, S., Herman, R. A., Nolan, C. L. & Wallen, K. 93. Avenanti, A., Bolognini, N., Maravita, A. & Aglioti, S. M. bases of empathic accuracy. Proc. Natl Acad. Sci. USA Men and women differ in amygdala response to visual Somatic and motor components of action simulation. 106, 11382–11387 (2009). sexual stimuli. Nature Neurosci. 7, 411–416 (2004). Curr. Biol. 17, 2129–2135 (2007). 112. Osborn, J. & Derbyshire, S. W. Pain sensation evoked by 75. Bufalari, I., Aprile, T., Avenanti, A., Di Russo, F. & A TMS study showing that the somatosensory and observing injury in others. Pain 148, 268–274 (2010). Aglioti, S. M. Empathy for pain and touch in the motor systems interact during the observation of This combination of psychological testing and fMRI human somatosensory cortex. Cereb. Cortex 17, extreme joint stretching. shows that a third of people literally share the pain 2553–2561 (2007). 94. Buccino, G. et al. Action observation activates of other people’s injuries on their own bodies and This EEG study used the timing of somatosensory premotor and parietal areas in a somatotopic links this feeling to vicarious activity in SI and/or SII. evoked potentials to show that higher stages of the manner: an fMRI study. Eur. J. Neurosci. 13, 113. Etzel, J. A., Gazzola, V. & Keysers, C. An introduction somatosensory cortex are modulated by the 400–404 (2001). to anatomical ROI-based fMRI classification analysis. observation of touch and pain, but BA3 is not. 95. Pierno, A. C. et al. Neurofunctional modulation of Brain Res. 1282, 114–125 (2009). 76. Allison, T., McCarthy, G. & Wood, C. C. The relationship brain regions by the observation of pointing and 114. Stoerig, P. & Cowey, A. Blindsight in man and monkey. between human long-latency somatosensory evoked grasping actions. Cereb. Cortex 19, 367–374 Brain 120, 535–559 (1997). potentials recorded from the cortical surface and from (2009). 115. Haggard, P., Christakou, A. & Serino, A. Viewing the the scalp. Electroencephalogr. Clin. Neurophysiol. 84, 96. Molnar-Szakacs, I., Kaplan, J., Greenfield, P. M. & body modulates tactile receptive fields. Exp. Brain 301–314 (1992). Iacoboni, M. Observing complex action sequences: Res. 180, 187–193 (2007). 77. Banissy, M. J. & Ward, J. Mirror-touch synesthesia is the role of the fronto-parietal mirror neuron system. 116. Botvinick, M. & Cohen, J. Rubber hands ‘feel’ touch linked with empathy. Nature Neurosci. 10, 815–816 Neuroimage 33, 923–935 (2006). that eyes see. Nature 391, 756 (1998). (2007). 97. Hasson, U., Nir, Y., Levy, I., Fuhrmann, G. & 117. Lenggenhager, B., Tadi, T., Metzinger, T. & Blanke, O. 78. Banissy, M. J., Kadosh, R. C., Maus, G. W., Walsh, V. & Malach, R. Intersubject synchronization of cortical Video ergo sum: manipulating bodily self- Ward, J. Prevalence, characteristics and a activity during natural vision. Science 303, consciousness. Science 317, 1096–1099 (2007). neurocognitive model of mirror-touch synaesthesia. 1634–1640 (2004). 118. Urban, P. P. et al. Different short-term modulation Exp. Brain Res. 198, 261–272 (2009). By analysing the degree of synchrony among the of cortical motor output to distal and proximal 79. Freund, H. J. Somatosensory and motor disturbances brain activity of many viewers of the same upper-limb muscles during painful sensory nerve in patients with parietal lobe lesions. Adv. Neurol. 93, western movie, this methodologically pioneering stimulation. Muscle Nerve 29, 663–669 (2004). 179–193 (2003). study shows that BA2 is active while the 119. Avenanti, A., Minio-Paluello, I., Bufalari, I. & 80. Hikosaka, O., Takanaka, M., Sakamoto, M. & hand–object interactions of other people are Aglioti, S. M. The pain of a model in the personality of Iwamura, Y. Deficits in manipulative behviors induced being observed. an onlooker: influence of state-reactivity and by local injections of muscimol in the first 98. Cross, E. S., Hamilton, A. F. & G ra f to n , S . T. Building a personality traits on embodied empathy for pain. somatosensory cortex of the conscious monkey. Brain motor simulation de novo: observation of dance by Neuroimage 44, 275–283 (2009). Res. 325, 375–380 (1985). dancers. Neuroimage 31, 1257–1267 (2006). 120. Minio-Paluello, I., Avenanti, A. & Aglioti, S. M. Left 81. Fabbri-Destro, M. & Rizzolatti, G. Mirror neurons and 99. Haslinger, B. et al. Transmodal sensorimotor networks hemisphere dominance in reading the sensory qualities mirror systems in monkeys and humans. Physiology during action observation in professional pianists. of others’ pain? Soc. Neurosci. 1, 320–333 (2006). (Bethesda) 23, 171–179 (2008). J. Cogn. Neurosci. 17, 282–293 (2005). 121. Avenanti, A., Minio-Paluello, I., Bufalari, I. & 82. Rizzolatti, G., Ferrari, P. F., Rozzi, S. & Fogassi, L. 100. Calvo-Merino, B., Grezes, J., Glaser, D. E., Aglioti, S. M. Stimulus-driven modulation of motor- The inferior parietal lobule: where action becomes Passingham, R. E. & Haggard, P. Seeing or doing? evoked potentials during observation of others’ pain. perception. Novartis Found. Symp. 270, 129–140 Influence of visual and motor familiarity in action Neuroimage 32, 316–324 (2006). (2006). observation. Curr. Biol. 16, 1905–1910 (2006). 122. Avenanti, A., Bueti, D., Galati, G. & Aglioti, S. M. 83. Rizzolatti, G. & Sinigaglia, C. Mirror neurons and motor 101. Calvo-Merino, B., Glaser, D. E., Grezes, J., Transcranial magnetic stimulation highlights the intentionality. Funct. Neurol. 22, 205–210 (2007). Passingham, R. E. & Haggard, P. Action observation sensorimotor side of empathy for pain. Nature 84. Rizzolatti, G. & Sinigaglia, C. The functional role of the and acquired motor skills: an fMRI study with expert Neurosci. 8, 955–960 (2005). parieto-frontal mirror circuit: interpretations and dancers. Cereb. Cortex 15, 1243–1249 (2005). misinterpretations. Nature Rev. Neurosci. 11, 102. Rizzolatti, G., Fabbri-Destro, M. & Cattaneo, L. Acknowledgements 264–274 (2010). Mirror neurons and their clinical relevance. Nature V.G. and C.K. conceived and wrote the Review, and J.H.K. 85. Rizzolatti, G. et al. Functional organization of inferior Clin. Pract. Neurol. 5, 24–34 (2009). wrote the anatomical section and provided comments on the area 6 in the macaque monkey. II. Area F5 and the 103. Flanagan, J. R., Vetter, P., Johansson, R. S. & manuscript. The work was supported by a Marie Curie control of distal movements. Exp. Brain Res. 71, Wolpert, D. M. Prediction precedes control in motor Excellence Grant of the European Commission, a VIDI grant 491–507 (1988). learning. Curr. Biol. 13, 146–150 (2003). frm the Dutch Science Foundation (NWO) to C.K. and VENI 86. Evangeliou, M. N., Raos, V., Galletti, C. & Savaki, H. E. 104. Miall, R. C. & Wolpert, D. M. Forward models for giant (NWO) to V.G. The authors wish to thank D. Arnstein, Functional imaging of the parietal cortex during action physiological motor control. Neural Netw. 9, A. Avenanti and L. Fadiga for comments, L. Cerliani and execution and observation. Cereb. Cortex 19, 1265–1279 (1996). S. Rozzi for discussions on the anatomical connections of the 624–639 (2009). 105. de Lussanet, M. H. et al. Interaction of visual hemifield somatosensory cortices and insula, L. Aziz-Zadeh, D. Dinstein 87. Raos, V., Evangeliou, M. N. & Savaki, H. E. and body view in biological motion perception. and L. Turella for their data on action observation and Observation of action: grasping with the mind’s hand. Eur. J. Neurosci. 27, 514–522 (2008). F. Filimon for re-examining it in relation to the location of Neuroimage 23, 193–201 (2004). 106. Hennenlotter, A. et al. A common neural basis for BA2. They would also like to thank S. Ebisch for providing the 88. Raos, V., Evangeliou, M. N. & Savaki, H. E. Mental receptive and expressive communication of pleasant lower half of figure 3 and A. Avenanti for the figure in Box 2. simulation of action in the service of action perception. facial affect. Neuroimage 26, 581–591 (2005). J. Neurosci. 27, 12675–12683 (2007). 107. Leslie, K. R., Johnson-Frey, S. H. & Grafton, S. T. Competing interests statement 89. Keysers, C. & Gazzola, V. Expanding the mirror: Functional imaging of face and hand imitation: The authors declare no competing financial interests. vicarious activity for actions, emotions, and towards a motor theory of empathy. Neuroimage 21, sensations. Curr. Opin. Neurobiol.19, 666–671 601–607 (2004). (2009). 108. Adolphs, R., Damasio, H., Tranel, D., Cooper, G. & FURTHER INFORMATION 90. Etzel, J. A., Gazzola, V. & Keysers, C. Testing simulation Damasio, A. R. A role for somatosensory cortices in Christian Keysers’ homepage: theory with cross-modal multivariate classification of the visual recognition of emotion as revealed by three- http://www.bcn-nic.nl/socialbrain.html fMRI data. PLoS ONE 3, e3690 (2008). dimensional lesion mapping. J. Neurosci. 20, The Juelich Histological Atlas: http://www.fmrib.ox.ac.uk/ Study showing how multi-voxel pattern 2683–2690 (2000). fsl/data/atlas-descriptions.html#jul classification can be used to examine whether 109. Pitcher, D., Garrido, L., Walsh, V. & Duchaine, B. C. The Statistical Parametric Mapping Anatomy Toolbox: the same representations are recruited during the Transcranial magnetic stimulation disrupts the http://www.fz-juelich.de/inm/inm1/spm_anatomy_toolbox execution of an action and the perception of other perception and embodiment of facial expressions. ALL LINKS ARE ACTIVE IN THE ONLINE PDF people’s actions. J. Neurosci. 28, 8929–8233 (2008).